Conrprrrers

& Strucrure.s,
Vol. 2. pp. 79-109. Pergamon Press 1972. Printedin Great Britain

PLASTIC ANALYSIS OF CASTELLATED BEAMS-I

INTERACTION OF MOMENT, SHEAR AND AXIAL FORCE

A. N. SHERBOUKNE~ and J. VAN OOSTROM$

Department of Civil Engineering, University of Waterloo, Ontario, Canada

Abstract-The Stress Resultant and Non-Local Theories of shear and moment interaction at fully plastic
sections are critically reviewed and practical lower bound relationships for shear, moment and axial force
interaction of plastic, rectangular and T-sections are derived. These relationships are used for developing
an incremental digital computer procedure for the analysis of castellated Beams in the elastic and plastic
ranges.

NOTATION
A area;
AF area of flange ;
AT area of direct stress due to axial force on T-section;
Aw area of web ;
AREA area of cross-section ;
b width of rectangular section or outstanding flange;
C stem height of T-section;
e distance from equal area axis;

5 distance to axial force region;

_f t-e;
h height ;
K,, Kz, K3 stress parameters ;
L length ;
M moment ;
MP plastic moment ;
MT reduced plastic moment in presence of axial force ;
m WM,;
axial force;
N

NP fully plastic axial force ;

NIN, ;
TEGRA;f uvh;
t flange thickness;
V shear force;
VP fully plastic shear force ;
v v/v,;

t Currently Visiting Professor, National Civil Engineering Laboratory, Lisbon, Portugal.

: Design Engineer, Dominion Bridge Co., Lachine, Quebec.

79
80 A. N. SHERBOURNE
and 1. VAN OOSTROM

web thickness ;
global axes;
member axes ;
distances to centroids;
cos-’ n;
arbitrary small quantities;
Lagrangian multipliers;
yield parameter ;
normal stress;
yield stress ;
shear stress;
yield stress in shear;
value of # at y=y,;
sin-‘(2T,,v/o,).

INTRODUCTION
THE TECHNIQUE of increasing the second moment of area of a rolled I-beam by converting
to a castellated beam has been known for many years in the shipbuilding and building
industries. Its increasing use was directly related to the perfection and acceptance of
welding techniques in general. The earliest tests in the United States were carried out by
Altfillisch et al. [l] in connection with the design of simple span highway bridges. In these
tests the measured deflections agreed well with an approximate elastic analysis using a
simplified Vierendeel truss analogue. Variations of this method have been used in analysis
and design of most castellated beams up to the present.
Gibson and Jenkins 123 presented a more refined model by taking into account the
bending stiffness of the various elements of the Vierendeel truss. Photoelastic investigations,
coupled with strain gauge and deflection measurements, showed results to within 15 per
cent of theoretical predictions. Kolosowski [3] further confirmed this theory by a beam
test and presented a second Vierendeel analogue based on the moment distribution method.
Gardner [4] improved the analysis by adding shear effects and narrowed the accuracy of
deflection predictions to 5 per cent. He also indicated that the error in neglecting direct
shear effects becomes too large to be ignored when the beam shortens to approximately
a dozen castellations of a standard configuration.
The current investigation attempts a plastic analysis of castellated beams by developing
a computer model which duplicates behaviour under incremental, proportional loading.
In this way, no particular collapse mechanism need be determined a priori, but areas of
plasticity may occur as the generalized forces reach limiting values on plastic interaction
surfaces for particular cross-sections of the beam. In consequence, it is important to
develop reliable interaction relationships between axial force, shear and moment in order
to ascertain whether a section has crossed the threshold from elasticity to plasticity.
Moment, shear and axial force interaction surfaces are developed for critical sections
of a castellated beam and any other I-section beam or girder with a hole in the web. The
sections considered specifically are the rectangular and T-section. The interaction curves
for these shapes are compared with that for the continuous I-section in order to obtain a
clearer unders~nding of the change deriving from a change in shape. Furthermore, inter-
action curves are plotted for a complete range of castellated beams in order to assess the
effect of changes in size and profile of the web elements and their attendant discontinuities.
 PlasticAnalysis of CasteiW%lBeams-I. Interactionof Moment, Shear and Axial Force 81

The particular range of beams chosen for this purpose consists of types 1, 3 and 5 of the
IPE 300 section as obtained in the tests and analysis of Halleux [5J. The cotiguration of
this beam is detailed in Appendix I.
In this paper an attempt is made to derive useful lower bound infraction surfaces
which can be applied in subsequent analysis of castellated beams. No corresponding upper
bounds are obtained so that the significance of the relationships cannot be assessed with
precision. However, one of the merits of this approach is that the results are conservative
and can, therefore, be used with safety.

STRESS RESULTANT THEORY-MOMENT/SHEAR INTERACTION

In a rigorous approach to the two dimensional problem of interaction between moment
and shear, it is necessary to satisfy both conditions of equilibrium and yield, respectively.

uX2+4v2r2X)’
<u 02 (2)

where o, is a normal stress, rsl a shear stress, o, is a yield stress in simple tension and v2
takes the value I or 314 depending upon the yield criterion, Tresca or von Mises, that is
invoked. The coordinate axes x, y are conventionally de&red.
The Stress Resultant Theory after Hodge [6] ignores local equilibrium and treats the
moment and shear forces as generalised variables at a section satisfying the yield criterion.
In consequence, it is an approximate theory. The generalised forces may be obtained by
integrating the stresses as follows:

M=
A
Yd-4 ; v=
s
A
r,,dA 9 (3)

Since Y is assumed to be a relative ~mum, 6 V vanishes to a first order and since M is

assumed to be constant, 6M vanishes for all variations of cr,. It follows that:

2vc5V+ISM=O
for all constants A.
The final values of M and V can be expressed in parametric form:

(4

dA. (5)

The interaction curve, M vs. V, for a particular section is obtained by eliminating 1

between equations (4) and (5) and integrating over the cross-section.
82 A. N. SHERBOURNE
and J. VAN OOSTROM

Figure 1 shows the profiles of the rectangular, tee and castellated beam sections,
respectively. The I-section is obtained by setting d=O in the dimensions of the castellated
section. For these sections the moment and shear expressions are given by:
(i) Rectangle

$=&[($)(y +l)i-sinh-‘($)I

’ - 2sin-1 Ah (6)
V, vh 0z
where

bho
I’,= bhz,=L
2v

RECTANGULAR

T-TYPE

CASTELLATED

FIG. 1. Typical beam sections.

 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 83

(ii) T-se&on
For this section, the only practical case considered is that in which the equal area
axis falis within the flange; this is realistic considering the usual dimensions of casteRated
beams. For this situation, the equations become:

M t
mz-= ~(2b+w){L(t-e)[12(t-e)2-i-1]*-sin-’[4t-e)]
M, A2AREA@, +J,)

+le(X2e2 -i-I)* +sinh- ‘( - le)> +w( - iie(L2e2+ l)* - sinh”( - ne)

-sir&-‘[&-e-c)])j

where

a,AREA
Vp=r,AREA=
2v

and 9 is a centroidal distance measured from the neutral axis.

(iii) Castellatedbeam
This section is integrated with respect to an axis dividing the total area symmetrically
yielding the following expressions for non-dimensional moment and shear:

“=~“~~A/[(26+w)(sinh-‘[~(~+c+S)]-sinh-’~l(d+e)lJ
P

+w(sinh”‘[k(d -t-c)]-sinh-‘(dd))j
84 A. N. !fh3RBOURNE and J. VANOosrtto~

(iv) kiect ion

In order to assess the vaiidity of the previous relationships, the nondimensional
moment and shear for this section is also obtained by setting d=O in the dimensioning of
the castellated section. The corresponding solutions agree with those given by Hodge f6]_

m=__=E
r.AtEAj [{2b+w}(~(c+?)[~“(c+t)‘+l]*-sinh-’[l(c+t)]
P 1

- Ac(~~c~+ 1)* -t-sinh- ‘(AC)}+ w(~c(~~c’ + l)* - sinh- i(k)}]

t~=F=&A[{Zb+wj{sinh-r[l(c+t)]-sinh-r(k)] +w(sinh-‘(lc))j . (9)

P

Interaction curves for the various shapes (rectangle, Z’, castellated, I} are plotted in
Fig. 2 for arbitrary values of the multiplier, 1 (1. The limiting values of 1 occur at the
extremes of the range when (M/M,)-+0 as (V/V,,)-*1 and (M/M,)-+1 as (V/V&-+0. The
castellated section is of the type as described in Appendix I, while the other shapes are
made up of its flange and web elements have the total depth of the T-portion. The curves
are obtained by varying the Lagrangian multiplier 1 from OGOO6to 3.4.

TYPE I SECTION TYPE 3 SECTION TYPE 5 SECTION

O.*O[S-T- TYPE

/4-I-TYPE
I I , / I I / I
000 020 040 060 080 loo 0.0 020 040 060 060 100 00 0.20 040 060 080 100

m =MIMP m=M/MP m=MlMP

FIG. 2. Shear-momentinteraction curves (stressresultant theory).

This approach recognizes the fact that there is no unique interaction curve relating V
and M, but that the relationship is a function of the geometry and loading of the entire
beam. However, the interaction curve for a cantilever loaded by an end load only can be
considered as a practical and very close engineering solution to the Vierendeel truss
behaviour of the members of a castellated beam. The rules for lower bound solutions are
followed: the yield condition is not violated and equilibrium is assured. Furthermore,
boundary conditions are also satisfied. When applying these rules as much of the section
as possible is required to attain the yield condition.
 Plastic Analysis of Castellated Eeams-I. Interaction of Moment. Shear and Axial Force 85

(i) Rectangle
The complete theory for moment and shear interaction in a rectangular cantilever
may be found in Drucker [S] and will not be repeated here. The relevant equations are:

for hlL<n/2, or V/V,s2/7r,

;=Z(L/h)‘[l -cos(Wl
P

$=(L/h)[l -cos(h/L)] (10)

P

for hlL>n/2, or V/Vpr2/n,

-$.=(4/n) 1- D/h)V/V,
P

$=(2/n)[l +(D/h)(nlZ- l)] . (11)

P

The geometry of the cross-section is the same as in Fig. 1, the length of this cantilever
is designated by L, and D is the distance between the normal stress zones for this high
shear range. To obtain the curve, h/L is varied between 0 and n/2 when equations (10)
apply, and D/h is varied between 0 and 1*Owhen equations (11) apply.

(ii) T-section
Figure 3 shows the geometry and assumed primary stress system for a T-type cantilever.
Normal stresses u,,, 6, and r,_ are taken to be zero. The equations of equilibrium of the
system are:

(12)
The yield criterion must not be violated at any point in the cantilever.

c,’ +4v2(rf, + 73 6 0;. (13)

In order to satisfy these equations of equilibrium and the yield criterion, three constants
Ki, K2 and Ks, whose absolute values must be less or equal to 1, are incorporated in the
equations for u,.

(a) Flange primary system

(14)
86 and J. VAN OOSTROM
A. N. SHERBOURNE

f
iEqual AM Axis

FIG. 3. T-section cantilever (geometry and primary stress distributions).

From equations (12) and (14),

a7,,laz= -ao,/ax= -(I/L)K,o,
and therefore,
5xz= [(b+ W/2-z)/L]K,a,. (15)

This satisfies the boundary conditions at z= w/2+b. It is sufficient to consider only the
area of positive axial stress, i.e. where ~20, because the equations in the complementary
zone will change in sign only. The yield criterion at the critical points in the flange, B and
C, must not he violated; the critical section occurs at x=L and z= w/2. From equation
(12)9
K?o;: +4vz(bZ/LZ)K:a; < f-r;
or
K, I 1/[1+(2vb/L)‘]*. (16)
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 87

(b) Web primary system

u, = (x/L)KAl

%, -0. (17)
From equations (12) and (17),

a7,,/az = - ao,/ax= - (1/L&r,.

Hence,

This satisfies the boundary conditions and equilibrium. The yield criterion at the critical
point in the web, x =L and y= e, must not be violated. From equation (13)

K;uf, +4vZ(cZ/LZ)K$3; < u;

or
K,S I/[1 +(2vc/L)]f. (1%

(c) Flange-web junction primary system

u,=(x/L)K,u,. (20)

Since the shear functions must be continuous in or&r to satisfy the equilibrium equations,

(21)

Let the yield criterion be satisfied at y = 0, z = w/2 and x = L. From equation (13), therefore,
at y=O,
7,=ue[(l-Kf)/4v2-(b2/L2)K:]*. (22)

The continuity of the vertical shear function, as well as the boundary conditions at
Y= - t + e, lead to the following:

for ~50,

t,..=(y +t -e)/(t-e)u,[(l -Ki)/4v2-(b’/L’)Kf]* (23)

for OSySe,

7w =a&1 - K3)/4v2-(b2/L2)K~]*+(y/e)((c/L)K2u,-a,[(1 - K:)/4v2

-(b2/L2)K:]*}. cw

The condition r_,/,, _, I zxyly cg must also be satisfied, i.e.

u,(c/L)K2 Q a,[(1 - K:)/4v2 -(b2/L2)K:]*. (25)

88 A.N. SHERBOURNE and J.VAN OOSTRGM

Equilibrium equation (12) must be satisfied throughout the junction :

aff~la~={il~)K~~~
ar,,jar = (~~/wL)K~~, .
For ys0,
a~,,/ay=(i/t-e)a,C(l-K:)/4~~-(b~/~~)K:]*,

For Olyle,
dr,,/ay = (l/e)~&/L)K, - [(l - K,2)/4v2- (b2/r;z)l;c:]*)
.

Hence,
KS/L-l-(Zb/wL)K, +[l/(t-e)][(l -K3)/4v2-(b2/Lz)Kf]*=0 (26)
and
K,/L+(2b/wL)K, +(lfe){(c/L)K2- [(l- K:)/4v2 -(b2/L2)KfJ*j =O. (27)

Thus far there are only two equations to solve for the three unknowns, K,, K2 and K3.
It is therefore necessary to obtain another equation to derive a unique solution. Ideally,
one would Iike to use both equations (16) and (19) to specify yield in both Aange and web,
but this is impossible. Equation (16) can therefore be maximized because then equation
(19) is not violated; the converse will not be true for any practical T-type section. Therefore,

K, = l/El +(2vS/Q2]*. (28)

Solving (26) and (27) simultaneously gives,

+(K:[(2b+e)/~L)~+(b/L)~]- 1/4v2) -0. (2%

Using the quadratic formula, the value of KJ is found. X2 can then be obtained from

K,=(L/c)([(l +>/4v2 -(b2/L2)Kf]*-(e/L)K, -(ZbelwL)K,j . (30)

If an imaginary value is obtained for KS, which occurs when the cantiiever becomes fairly
short, the previous equations no longer apply. At that point, Kj has become quite small
and can conveniently be set to zero. Equation (28) must be discarded and replaced with
inequalities (16) and (19), which will be satisfied automatically. Equations (26) and (27)
can then be solved simultaneously to obtain:

KS =L/[2v6([2(t-e)fw]2+ Ij*j (31)

K, =(L/c){[l/4v2 -(b2/L2)Kf]*-2be/wL}. (32)

 Plastic Analysis of Castellatcd Beams-I. Interaction of Moment, Shear and Axial Force 89

Condition (25) is found to be satisfied except for the extremely short range of cantilevers.
When this equation is violated, one must drop the junction yield condition. Equations (26)
and (27) then become:

leading to,
K 1= - cK2w/2bt (33)
and
‘5NA
= (t - e)/t(c/L)K,a, . (34)

Since t,,=(c/L)K2 CT
O,at y= e, the critical point for violation of the yield criterion becomes
y= e. From equation (13), then,

K 1= -U[4v%2( 1 +4t2/w2)] (35)

K 2= -2btKJcw. (36)

For all but very short cantilevers, one can integrate over the cross-section to obtain

V = wao((cK2/2L)(c +e) + t/2[1- K:)/4v2 -(b’/L’>K:]*) . (37)

For very short cantilevers,

V= wa,,[(cK2/2L)(c + t)] . (38)

The nondimensional shear and moment for the primary stress system, u1 and ml respec-
tively, then become:
u, = V2v/a,AREA

m,=2VL/[a,AREA(J, +J2)]. (39)

It is observed that, due to strict adherence to equilibrium and boundary conditions as well
as non-violation of the yield condition, in no case does yielding occur anywhere in the
web; for shorter cantilevers it is not attained in the flanges either. Therefore, secondary
stress systems, which in themselves obey the rules for a lower bound solution, may be
superimposed on the web and the flanges, when applicable, so that they will reach yield
at critical points.
(d) Web seconukry system. In this stress system t,, is again equal to zero. The
equilibrium equations are:
aaJax + a7,,/ay = 0
and
a7,,lax=o.
90 A. N. SHERBOURNE and J. VAN OOSTROM

Therefore,

6, = - Wr,,ldy) . (40)
At x=L,

6, = - L(dr,,/d_v) (41)

The secondary bending moment will superimpose a negative axial stress on the half of the
web closer to the flange and a positive axial stress on the remainder. The critical point for
yielding will be at y = e + c/2. At that point, the primary stress system has cr, = K,aO and
xy = (c/2L)&%. Therefore, at that point, from equation (13).

(42)

This equation can be solved directly on the computer using a Runge-Kutta-Gill library
sub-routine, but it is found that the accuracy deteriorates rapidly as the cantilever becomes
shorter. The mathematical procedure employed in the solution necessitates the following
substitutions :

K,=n

leading to an interaction curve of the form:

(no0- L(dr:,/dy)]’ +4v2r;3 = 0; (43)

from which u2 and m2 can be obtained [7]

u2 = (wv/AREA)[TEGRAL - (c2/ L)K,]

m2 =[wL/AREA(jj, +j,)][TEGRAL-(c’/L)K,] (44)

where TEGRAL is defined as the quantity (uvh) of the rectangular section considered.
It is found, however, that in this case both the numerical and analytical solutions give
inaccurate answers for u2 and m, when L becomes quite small, due to an inherent instability
of the transcendental relationship between v, m and axial force for the rectangular section.
If a simple interaction relationship between moment, shear and axial force is assumed of
the type:

m’ + (B’)~+ ( v’)4/[ 1 - (n’)‘] = 1 (45)

the revised expressions become :

uy = (cw/AREA)[o’ - (c/2L)K,]

m2=c2wm’/[2AREA(jj, +jj2)]. (46)

 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 91

(e) Flange secondary system. At the point at which K3 is set to zero, Kl is found to
be well below its limiting value required from equation (16) to cause yielding. Therefore
a secondary stress system is introduced in the flange over this short length, high shear
region. The equilibrium equations are once again:

a0,lax
+ a7,,/ay
=0
and
dTxylaY = 0.
Therefore,
O, = - x(dr,,/dy) .
When x=L,
0, = -LW,ldy) .
In the primary stress system ux= K,a, and zxz =(b/L)K,a, at the critical point B. The
yield criterion (13) is just satisfied when

[Kl% -L(dr,/dy)]’ +4v2[7;,,+(b2/L2)K;a@ =a;.

Equation (47) differs slightly from equation (42) but, for L approaching the section depth,
the numerical solution on the computer again becomes inaccurate. An analytical solution
in this case is more satisfactory and the following arbitrary substitutions are made to
reduce it to solvable form:

ao”= ae( 1- 4v2K:bZ/LZ)*

n=K,/(l-4v2K:b2/L2)+.

Equation (47) then becomes

[mr;; -L(dr,/dy)]’ +4~~7;,,=0;;~ (48)

leading to solutions as follows:

u3 = [2b TEGRAL(1 -4v2K:b2/L2)*]/AREA

m3 = [2bLTEGARYl -4v2K:b2/L2)*]/[v AREA@, +J2)] (49)

where u1 and m3 are the non-dimensional shear and moment respectively of the secondary
stress system in the T-section flange [7].
The complete interaction relationship is then obtained by adding the component
systems.
v=o,+u,+u,

m=ml +m,+m,. (50)

92 and
A. N. SHERBOURNE J. VAN Oos~~ou

The resulting interaction curves for the total non-dimensional shear and moment as well
as the component stress system for the typical range of T-sections may be seen in Fig. 4
and in the representative values as given in Table 1. Even though only three points in the
section are at yield most of the other zones are fairly close to it especially for shorter
cantilever lengths. A slightly better lower bound may be obtained using trigonometric
functions for the primary system but the improvement is not worth the large increase in
complexity.

TYPE 1 SECTION TYPE 3 SECTION TYPE 5 SECTION

j 1 / / j

m= M/ MP m=M/MP m =M/MP

FIG.4. Shear-momentinteraction for T-section (non-local thecry and Trcsca yield criterion).

TABLE 1. Moment-shearinteraction for type 1 T-section

(non-local theory and Tresca yield criterion)

Cantilever
length
(mm.) m 1266 400 300 225 169 95 53 30 11 0

kl 1GOO 0.994 O-942 0901 O-776 O-582 0.327 0.184 0.104 0.038 0.000

k2 1 .OOO O-993 0.931 0.882 0.739 0.554 0.312 0.175 0.099 0.036 OWO

k3 1 GO0 0.972 O-712 0.456 OGOO O+lOO O-000 O*ooO OGIO OGOO OWIO

ml 1GOO 0.993 0.931 0.882 0.739 0.554 0.312 0.175 0.099 0.036 OGIO
m2 OWO OWO O@OO O*OOO0.066 0.177 0.274 O-304 0.312 O-295 OWO

m3 O+OO 0.004 0.033 0.055 O-112 O-170 0.222 0.235 0.217 0.118 0.000

01 0~000 0.009 0.028 0.035 0.039 0.039 0.039 0.039 0.039 0.039 0.04
02 O-000 O-o00 OWO 0.000 0.003 O-012 O-034 O-067 O-112 0.315 0.83

03 0000 0.000 O*ooO 0000 O-000 0.002 O-018 0.042 0.075 0.116 O-13

m I.000 0.996 O-964 0.937 0.917 0902 0.808 0.714 0.627 0449 OGIO

L’ OGIO OGO9 O-029 0.037 0.042 O-059 0.091 O-148 0.236 0.470 1GIO
 Plastic Analysisof CastehatedBeams-I. Interaction of Moment, Shear and Axial Force 93

(iii) I-section
The lower bound solution based on a linear stress distribution can be found in Neal
[9]. The cross-section geometry is that for the castellated section of Fig. 1 with d=O;
L is the length of the cantilever. Constants K,, K, and K3 are again incorporated into the
axial stresses of the flange, web and junction respectively. The values of these constants
can be stated directly in nondimensional parameters:

K1= l/(1 +B)f

K2=[-DE+(l+D’-E2)*]/(1+D2)

K,=[-AC+(l+C2-A2)*]/[(1+C2)(1+B2)*].

When K3 is imaginary, i.e. L/2c<v(a2 -y2)*,

K, = 1/(A2+B2)+

K,=O.

The interaction relationship is based on the following equations:

1)1 = w -w*w +o- w(c/Jw2v(c + 0

2[wc + t(2b + w)]

m, = wL[(l -K;)*(2c +t)/2v-(cp&(c +t)]

(51)
t(2b + w)(zc + t) +c2w

When L/2c > v(a2 - y2)+, u2 = 0 and m2 = 0. When L/2c < v(a2 - y2)*, the following equations
lead to the secondary interaction relationship:

a=cos-‘n=cos-1[K,/(1-K~B2)*]
h=2t
+=vh/L-cotatanh-‘(cotatari+)
tXY= (ab/2v)sin 4
dy = (L/2v)cos 4 d4/(cos a f cos 4)
JO
V=2b Qy
s -hj2

=(2bInb/2v2) 1 [sin t#~

cos’ &(cos2 a-cos’ #)]dQ) (52)
s
where
ob = q,( 1- KfB’)* .
94 A. N. SHERMXJRNE and J. VAN OOSTROM

Therefore,

u2= bL(1 -K:B2)*{1 -cos 1+9

+2coscl[tanh-‘(coscrsec$)-tanh-‘(cosu)]}
v[wc + t(2b + w)]

mz = bL2( 1- K:B’)*{l - cos $ +2 cos cl[tanh- ‘(cos CIset rl/)- tanh- ‘(cos a)]}
(53)
v2[t(2b+w)(2c+t)+c2w]

Then
v=v,i-0,

m=m, -km,. (541

(iv) Castellated section

One cannot use the Non-Local Theory with any justification for the whole castellated
beam unless the entire section, including cut-outs, is taken into account. One of the
equations of equilibrium states that dr,,/dx=O; this cannot easily be satisfied. Doing a
proper analysis, then, would be extremely complex if approached this way and thus, the
interaction curves for a castellated section, for purposes of comparison with other shapes
are not derived.
Shear-moment interaction curves are plotted in Fig. 5 for three representative shapes
(rectangle, T, I) based upon the theories outlined above.

TYPE I SECTION TYPE 3 SECTION TYPE 5 SECTION

OQOL I /
00 0.25 O&3 075 H)o 00 0.25 OS0 07s loo 00 025 050 0.75 PO0

m=M/MP m=M/MP m =t*/MP

FIG. 5. Shear-moment interaction curves (non-local theory and Tresca yield criterion).

COlUPARISON OF THEORIES
The results of these two theories may best be seen by comparing the curves of Fig. 2
with those of Fig. 5. It is observed that the shapes of the interaction curves for the rectan-
gular section are somewhat similar, while those of the I-type and T-type sections are vastly
different. The differences in results may be attributed to three reasons:
(i) The Stress Resultant Theory ignores local equilibrium and is therefore statically
inadmissible. At best it provides an estimate in contrast to the Non-Local Theory which
is a more correct lower bound solution.
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 95

(ii) In assuming a unique interaction curve for shear and moment for a particular
cross-section one is using normal beam theory which assumes that moment deformations
are more important than shear deformations, which are usually ignored. An element
subjected to moment tends to deform very differently from the same element subjected to
shear only. The combination of these two actions causes a mismatch, that is, another
equivalent action-deformation system by which some parts of the cross-section are affected
by shear while others are not. This mismatch would not really exist in a beam, but neigh-
bouring sections, instead, would be pulled out of shape by the resulting action-deformation
system. The actual stress distribution, therefore, depends on the entire beam and loading
pattern. In elastic design this mismatch is of second order magnitude because there is a
gradual transition from an area of maximum moment to zero moment. At a plastic hinge,
however, deformation is localized and there is an abrupt transition between deforming and
non-deforming material. However, for pure bending there is no problem of mismatch;
the problem becomes acute when the cross-section changes greatly and shear occurs, and
hence local criteria cannot apply for such a rapidly changing cross-section. For the
rectangular cross-section the Stress Resultant Theory gives a fair estimate, while it errs
considerably for other sections. For problems of shell analysis it is probably satisfactory
to use this type of theory but in the current situation it is not.
The cantilever analogy used in Non-Local Theory is sufficiently close to the situation
of the elements of a castellated beam to be used directly in the analysis.
(iii) The third reason has already been partially stated in the mismatch phenomenon
but can be looked at in a different way. Taking the T-Section as an example one can see
that for long lengths the flange becomes plastic when the web still retains capacity to
carry stresses. Therefore, part of the web will be in equilibrium with the plastic flange
while full plasticity is obtained by the addition of an exclusive web equilibrium system.
In other words the web is too stiff for the flange to have a single system. This can be seen
intuitively by inspecting Fig. 6 which shows the relative inertias of the flange and webs for
the I.P.E. range of castellated beams. It is seen that the web is much stiffer than the flange
for the complete range. For shorter cantilevers the primary equilibrium system is unable,

1.76 I IO’
IorJ x to*
Ss2bZrO’
316 xlO"

1.76x IO'

loOxlo'

562r104

3BrlO*
176X04

ICOXIO.

562x10'

316x0=
200 400 600 600 ICWJ
WEB HEIGHT, mm

FIG. 6. Moment of inertia components of T-section. (Series 2, castellated beams.)

96 A. N. SHERBOURNE and J. VAN OOSTROM

by itself, to cause plasticity in the flange and a secondary system containing vertical shear
is imposed within the flange only. There are three distinct systems at work, then, the
overall or primary system, the web system, and the flange system. One can see that there
is some intuitive justification for this by looking at the three different inertias in Fig. 6 and
visualizing each as acting as a partially independent unit. It is this physical phenomenon
which the Stress Resultant Theory ignores. It assumes full vertical shear carrying capacity
in the flanges while the other theory, as well as experiment, shows that the flanges cannot
carry any shear until the cantilever length becomes quite short. The Non-Local Theory
for the I-section is found to compare quite well to bounds given in Neal [9]. This phe-
nomenon explains the vast difference in the results of the two theories for the T and I-type
sections. Therefore, it is felt that one can only justifiably use Non-Local Theory having
conclusively proven that the Stress Resultant Theory is inadequate. In a later section
lower bounds for shear, moment and axial force interaction will be developed using Non-
Local Theory so that these may subsequently be used in the incremental plastic analysis
of castellated beams.

MOMENT/AXIAL FORCE INTERACTION

In order to obtain complete moment, shear and axial force interaction relationships
it is of basic importance to study the effect of axial forces on the plastic moments of various
cross-sections. For doubly symmetric sections the bending moment, M, will cause an
anti-symmetric distribution while axial force N will cause a symmetric distribution about
the axis of symmetry leading to a net displacement of the neutral axis as shown in Fig. 7.
The effect of axial force on the plastic bending moment of sections with a single axis of
symmetry in the plane of flexure is discussed by Eickhoff [lo], who also chose the equal
area axis about which M and N could act. A general interaction relationship, applicable
to both singly and doubly symmetrical sections, can therefore be developed.
Let MT be the reduction in moment due to the axial force which acts over an area A,..
Therefore,
M=lu,-MT
N=A+,
m=M/M,=l-MT/Mp=l-AT(~,+~Z)/A(_~l+Y~)
n=N/N,=ATa,/Aa,~A,/A. (55)
Therefore,
m= 1 -n6, +~2Y(J1 +YJ (56)

where it and T2 are distances to the centroids of the half axial force areas above and below
the equal area axis respectively; jji and jj2 are distances to the centroids of half of the
total area above and below the equal area axis respectively.

(i) Rectangular section

The geometry of the rectangular section may be found in Fig. 7(c). Because of double
symmetry, expression (56) becomes :

m= 1 -n;,/y2
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 97

FIG. 7. Axial force and moment stress distributions for various sections at full plasticity.

and
n = N/N,, = Zeb/hb = 2g/h
y1 = h/4

y’, =e/2,
Therefore,
Jl/p,=2e/h=n
and the equation of nondimensional moment/axial force interaction becomes:
m=l-2. (571
98 A. N. SHERBOURNEand J. VAN OOSTROM

(ii) T-section
The geometry of the T-type section may be found in Fig. 7(d). For simplicity it is
best to use expression (56) as it stands and not to redefine it in non-dimensional terms.
For the I.P.E. range of beams considered the equal area axis will always lie in the flange.
yl=(t-e)/2
L1 =A,/(8b +4w)
fz _(2b + w)e2/2 +(e +c/2)cw
e(2b+w)+cw ’

There are two possible cases for p2 depending on whether all of the axial force area, Ar,
lies in the flange or not. If Ar/(46+2w)>e, it will be partially into the web area. If
A,/(Jb+2w)<e, J2 =A,/(8b+4w)
A,/(46+2w)>e, jj2 =(l/Ar)[(26+ w)e’+ ivf’+2wfi]
where
f=[A,/2-(2b+w)e]/w
(iii) I-section
The moment/axial force interaction for the I-section is obtained by substitution into
equation (56). The expressions were also derived in Neal [l I]. Figure 7(e) shows the
geometry of the cross-section.
If n< A,/A,

1’
1
m=l-?a2
(l-A//A)[l-w/(2b+w)]
If n2 A,/A,

1’
n2-[l -w/(2b+w)](n-A,/A)’
m=l-
1 - (A,/A)'[l - w/(2b + w)]
Beedle [12] found good agreement with this theory in a series of column tests.
The curves of the various cross-sections for the I.P.E. range of beams are shown in
Fig. 8. It is of significance to note that the plastic moment of a T-section decreases very
little until the axial force becomes quite large.
TYPE 1 SECTION TYPE 3 SECTION TYPE 5 SECTION

I
0001
O-0 020 040 c-60 a20 04G 055 000 FGG 020 G-40 0 cd SiBG / i’0

Ill = M(b+p m=M/Mp m = rd;blp

FIG. 8. Moment-axial force interaction at full plasticity.
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 99

NON-LOCAL THEORY-MOMENT/SHEAR/AXIAL FORCE INTERACTION

(i) Rectangular sectiozz
The problem of the reduction of the fully plastic moment of a rectangular cantilever
due to normal and shear forces was solved by Neal [13]. As the results of this theory are
more or less directly applicable for the vertical members of the castellated beam it is
reviewed briefly.
Equilibrium equation (1) states that
h,.ax=-a7,,/ay. (W

It is assumed that the axial force N is uniformly distributed over the cross-section so that,
for x=0,
o, = N/bh = naO.

Integrating equation (58) for 01x< L, leads to

6, = - xdr,,/dy + nao . (5%

The yield criterion, equation (2) states that

a: +4v2r$=ai.

At the critical point, where x=L, it follows that

(no0 - Mr,,/dy)2 +4v2rz, = a$. WI

This equation may be put in a more convenient form by making the following substitu-
tions :
f ._ = a0 sin &/2v

n = cosa (OSaS71/2)
giving
(L/2v)cos&coW*cos~. (61)
dY
Upon integration, this leads to

(2v/L)(h/2+y)=4-2cotatanh-l

(2v/L)(h/2-y)=4+2cotatanh-’ ;y,<ych/,. (62)

At y= fh, T,~ -0. At y=y,, I$=$. Therefore,

$=++cotatanh-‘(cotatan$)

2Yo_ tanh- ‘(cosec a sin $)

(63)
h-II/ tan a+tanh-‘(cot atan $1’
100 and J. VAN O~STROM
A. N. SHERKHJRNE

The solution is obtained by specifying n, which gives cx. One then solves for li/ in
equation (63) for a particular value of L. It is noted that O<$<CZ.
Now
h/2
V=b cx,dy
s -h/z

For L/vh > 2/q

u = (2v/ho,)
s hi2

-h/2
rxydy.

) sin+cos’d,
V=
I 44
0 cos2 4 -cos2 Cx

(64)

For L/vh I 2/x

v=
s IL sin 4 cos2 Q,
o cos2 4 - cos2 u

where y,, is first found from equation (63).

d$ + 2y,lh (65)

Also
m=M/M,=Zu(L/vh). (66)

A straightforward solution of equation (64) shows that, as L becomes smaller, the

solution for u becomes more inaccurate; therefore the integral forms of the equations
were used to find the values of u in the computer programme. The fact that this function
is basically unstable for short lengths cannot be ignored, and, for very short lengths, no
reliable answers can be found. The instability of the function is due to very small and
similar numbers being subtracted or divided in the range L< h. This can be physically
interpreted as a possible limit to the applicability of St. Venant’s Principle.
Another instability of this function occurs when n is very small, which brings a close
to n/2. Then, for L 2 2vhln,

JI=vh/L.

When L 5 2vh/x, the following procedure is necessary. Let c(= n/2 - 8, and rl/= n/2 - p-y.
From equation (63)

vh,/L = n/2 + (p/2)ln( 1 + 2/3/y).

Solving for y,

(67)
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 101

Because the value of y is extremely small, it cannot be found explicitly due to computer
limitations. Equation (67) is therefore substituted for y in equation (63).

tanh- ‘(COW a sin I++)

$tana+tanh-‘(cotatari+)

(2vh-aL)/2L/?-In/l
(68)
(n/2-~)tan(n/2-/?)+(2vh-rL)/2L/3 ’

In (68) y has been eliminated. An arbitrary small value of /I must be chosen to obtain a
solution. Equation (65) may then be used to obtain a value for V.
The results may be seen in Fig. 9. The kink in the curve of n=O is due to the transfer
to the approximate solution involving equation (68). At that point both solutions tend to
be inaccurate. The curves for very short lengths were linearized on the graph because no
accurate solutions could be obtained due to the instability of the transendental function
for u. Therefore, for practical moment, shear and axial force interaction conditions for a
rectangular section, another approach is used. It is found that a simple algebraic equation
gives a good approximation to the lower bound interaction relation for m, a and n. The
discrepancy never exceeds 5 per cent and is usually much less. The error is greatest in the
intermediate range of cantilever lengths where length L is several times the beam depth h.
This expression takes the form

m+n’+u’/(l--n’)=l (69)

and is used whenever possible because of its simplicity. The results of this equation are
shown in Fig. 10 and are seen to parallel closely the curves of Fig. 9.

0
> 060
5
8‘

>

m = M/MP

FIG. 9. Shear/moment/axial force interaction curves for rectangular section. (Treaca yield criterion.)
102 A. N. SEERBOURNEand J. VAN OOSTROM

FIG. 10. Shear/moment/axial force interaction curves for rectangular section [empirical equation (6911.

(ii) T-section
(a) Analytical solution. The approach used is similar to that for shear and moment
interaction alone. The geometry and stress distributions are shown in Fig. 11, the equal
area axis is always considered to be in the flange and the normal force is considered to be
uniformly distributed at the free end of the cantilever and concentrated around the equal
area axis at the fixed end. The case in which the axial force is of the same sign as the stress
at the upper surface of the flange will be analyzed here. Again, the final interaction relation-
ship will be in non-dimensional terms, where n= N/N,, v= V/VP and m= M/M, and the
equations of equilibrium (12) and the yield criterion (13) again apply.
The conditions in the various elements of the beam will now be investigated in turn:
Web
flX= [n + (Kz - n)x/L]ao

z xz =o.
From equation (12), therefore,

giving
r,,=a,(K,-n)(c+e-?,)/L.

The yield criterion, equation (13), must not be violated.

Ki +4v2(cz/L2)(K2 -n)” < 1

or
K < 4nvZc2/L2 +[(4nv2c2/L2)2 -(l f4v2c2/L2)(4vZc2n2/L2 - l)]+
2. (70)
1 +4v2c2JL2
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 103

EQUAL Am AXIS
_-.

AREA @

=x = [n +(K,-“)x/L]q,

II ~~~~ (Kr-n)q,(c+e-yVL

1
Q) AREA @

Qx =[“+(K,-nh/L]co
~~~=(K,-n)u~(b+w/2-z)/L
AREA @

o,, =[n-1 K,+n)x/~]q,

rxz =-(K, +n)q, f bcwh?-r)/L
AREA @ AREA @

Qx = [n-(K~+nh/L]c, ux =[“+(Ks-n)x/L]c,
TX2 =-(2z/wNb/L)(K,+n)~~ rXx =(Kt-dq,( b/L)(Pz/w)
y+t/i+x1-c
=xy = Qo1 ~,,=~~{[fJ-Kf)/4~-(b~/~)(K,-“~]~
t (I+X)_e I
+(Y-XtV(e-Xt)[c/L lb-n)-(( I-Kf)/49
(I-K;2V4v2-(b2/f 1(K,+nf/@j
-(b2/L2)(K,-n)2f12])

FIG. 11. T-section (geometryand primary stresses).

Flange : y 2 At
u, = [n+ (K, - n)x/L]ao

?XY =o

7xz =a,(K, -?l)(b+ w/2 -2)/L. (70

Equation (71) satisfies the boundary conditions and the equation of equilibrium. The
yield criterion must not be violated at the critical point, when x=L and z= w/2. At that
point crx=Kl~,, and oxI= bo,(K, -n)/L. Applying equation (13) gives

K <4nvZb2/L2 +[(4nv2b2/L2)-(1 +4v2b2/L2)(4v2b2n2/L2 - l)]*

11 (72)
1 +4v2b2/L2
 A. N. SHERBOURNE
and J. VAHC~XTROM

5x=[n-(fR, +nkxlL]a,

T
XY=O

T xz= - a,(K, + n)(b + w/2 -z,/ L .

The yield criterion must not be violated at x =L and t= w/2, where a,= - K,o, and
Txz= - baO(K, + n)/L. Therefore,

K < -4nv2b2/LZ +[(4n~~b~/L~)~-(f +4vZb2/L2)(4v2b’n2/L2 - I)]*

1% (73)
l-k4v%2/L’

Junction: y2&

ax=[n+(K, -n)x/LJo,

z,, = 60( Ki - n)2zb/ WL.

If the yield condition is just satisfied at the critical point z= w/2, x =I, and y = it,

rxr = [CQ(1 - K:)/4v2 -(K, - n)2b2/L2]*. (74)

Therefore, for 3.t 5 _v5 e,

?xy=u*[[(l -K3)/4v2-(K, -n)‘b2/L21f ~(~-~t)/(e-~)~(K2-ff)c~L

- [(l - K3)/4v2 -(K, - n)‘b2/L2]*jjj

Junction : y< At

a,=[n-(K;+n)x/l]a,

T,, = - o&C1 -tn)szb/wL .

At z=w/2, x=L and y=If,

zx,,=50[(1+K~2)/4v2-(K, +n)2b2/L2]3. (7.5)

Equations (74) and (75) should be identicai, because there can only be one value of
t,, at y = It. Therefore,

Kj=[K$-8K,nv2b2/L2]f. (76)

For (e-t)lysii;t,

T~,=Q~[{[Y+~(~ -f-A)-e]/[t(l i-L)-e]}((l- K;*)/4v2-(K1 -~-n)~b~/L~}*j.

 Plastic Analysis of Castellated Beams-L Interaction of Moment, Shear and Axial Force 105

Applying the equations of equilib~um throughout the junction gives

(K,-n)/L-l-(Kr -n)2b/wL+[l/(e-Ir)]{(K,-n)c/L-[(l -K$/4v2

-(K, -n)2bzf12J*} =O (77)

and

-(K, fn)2b2/L2J+=0. (78)

i, may be determined from a simple geometric relationship:

. -nAREA+K,wc+Ki(46e-2bt)-Kjwe-K;(we-wt)
.. (79)
4bK~r+wf(K~ +I(>)

For given values of L and n, values of constants K,, K2, K,, K; and rZcan be obtained
using equations (73), which is maximized, (76X79), subject to conditions (70) and (72)
which are found to be automatically satisfied in the applicable range of L and n. The
equations in this form are only suitable for a relatively small axial force and relatively
low shear, the latter in turn implying a relatively long length of cantilever. The four con-
stants, K,, Kj, Ks and K; must be greater than n, and the limit of app~~bility of the
equations occurs usually when KS violates this condition, Using these constants, non-
dimensional terms u and m may be determined by integrating across the sectional area and
dividing by VP and Mp, respectively.

v=2vw/AREA((K2 -n)c2/2L+[(e-b)/2][(1 -K:)/4v2 -(Kr - n)2b2/L2]f

- e]}[( 1- K;‘)/4v2 -(K, +r~)~b~/L~]*)

m= t&/v@, + y2). WV

This solution is similar to the primary stress system solution for the T-section subject
to shear and moment only. It is plotted in Fig. 12. The method can be extended to shorter
cantilevers and higher axial force by making the junction direct stresses equal to zero,
superimposing secondary stress systems and distributing the axial force among these as
well. The complications, however, become quite severe and the resulting solutions become
less and less realistic. The interaction surface would consist of a patchwork of solutions,
each one being valid only over a very short range. It may be desirable to obtain all these
solutions but it is felt that this would be beyond the scope of the present study and would
not really achieve the objective of the current study which is the plastic analysis of castellated
beams. This is especially true when an approximate solution, as described in the next
section, can be relatively easily obtained.
106 A. N. SHERBOURNE and J. VAN OOSTROM

TYPE I SECTION TYPE 3 SECTION TYPE 5 SECTION

y-- ‘OOj 1 I I 1 - lOOi n.rx I

I 025 050
m- M/MP m=M/MP m = M/MP

FIG. 12. Partial plot of analytical solution for shear, moment and axial force interaction in T-section.

(b) Geornetricul solution. In earlier sections the non-dimensional relationships between

shear/moment and moment/axial force for a T-section in the plastic condition were obtained.
These two curves are the intercepts in the m-u and m-n planes respectively of the three
dimensional m-u-/z surface produced by the total interaction relationship. An interaction
curve for v and n will now be developed so that the edge conditions of the total interaction
curve are completely defined. From these three boundaries the remainder of the surface
can then be generated.

Shear/axial force interaction

This condition can only occur in a member of infinitesimally short length. The equation
of equilibrium (1) must still be satisfied. Because direct stress is due to the axial force
only, do,/~3x=O, and therefore, dr,,/dy =O. This would be in violation of boundary con-
ditions for a member of any finite length.
The yield criterion (2) is the other condition to be used.
A variational approach, similar to that used earlier is considered suitable in this case.

N= o,dA
s .4

V= s,,dA=(1/2v) (a;-a2,)+dA
s A j A

6N = Ga,rdA
i‘ A

A - [aX/(ai - az)*]Ga,dA . (81)

Since V is assumed to be a relative maximum, by- just satisfying the yield criterion,
6V vanishes. Since N is a constant in this case, 6N vanishes for all 6a. Therefore,

2vdV+WN=O

or .
[ - a,/(az - a;)* + A]So,GA = 0 .
”! .4
 Plastic Analysis of Castellated Beams-I. Interaction of Moment, Shear and Axial Force 107

This leads to
b, = na,/( 1+ n2)* = <60

where r is defined as a new Lagrangian (Euler) multiplier as a function of Iz and equal to

A(1+12)+. This results in

IV=&, dA = {Aa,
sA

V = [a,(1 - c2)*/2v] dA = Aa,(l - r2)+/2v.

s A
Since
N,=Aa,

VP = AT, = Aa,/2v.

then,
n=N/Np=(

u= V/V,-(1 -c2)+ (83)

resulting in
D=(l -n2)*
or
u2+n2=l. 634)
This is recognized as the equation of a circle.

Development qf interaction planes

Using the three curves as plane intercepts cannot provide a unique surface. However,
by using the shape of the m-u curve and using the other two curves as end points, as
illustrated in Fig. 13, a credible interaction surface is generated. This is done by approxi-
mating the m-v curve by a polynomial which is then scaled to fit between the end points

n m

FIG. 13. Geometricai interaction surface (lines of constant n are similar with n=O-0).
108 A. N. SHERBOURNE and J. VAN O~STROM

which are deteimined from the other curves. A further approximation considers the form
of all lines for finite n as being identical (but scaled down) and typified by n=O. This
appears justified by inspection of Figs. 9 and 10 pertaining to the rectangular section.
Various m-u relationships, for contours of n are plotted in Fig. 14 for the range of I.P.E.
T-sections.

d/I/h=”

FIG.14. Approximate shear, moment and axial force interactions (geometrical solution).
 Plastic Analysis of Castehated Beams-I. Interaction of Moment, Shear and Axial Force 109

This surface for a T-section is used in the plastic analysis of castellated beams. Usually
the problem consists in checking to determine whether the moment at full plasticity has
been reached for a given n and II. It is true that the plastic moment-shear interaction
curve is related to a particular length L at each point. However, since this is only a lower
bound and therefore not the actual solution, the actual length of the cantilever becomes
somewhat longer than L used in the interaction relationship. The difference depends
upon the closeness of the lower bound to the true solution. Since there is no way of
measuring this beforehand, L cannot be specified as an input parameter and will therefore
not be used directly for the interaction relationship in the castellated beam solution.

I-section
The solution of the nr, u and n interaction relationship for part of the range may be
found in Neal [14].

CONCLUSIONS
Simple moment, shear and axial force interaction relationships have been derived
based upon lower bound, equilibrium analyses of the stress systems prevailing in rectangular
and T-sections. The purpose of this investigation is to provide convenient measures for
testing local plasticity in the various zones of a castellated beam loaded up to the point of
plastic collapse. Since these beams are usually analysed using Virendeel Truss Analogues,
it is important to be able to assess the individual limit behaviours of the vertical and chord
elements which are believed to approximate members of rectangular and T-sections
respectively.

AcknowZe&ements-This work was carried out in the Department of Civil Engineering of the University
of Waterloo under National Research Council of Canada Grant No. A-1582 to one of the authors (A.N.S.)
and an Ontario Graduate Fellowship to the other (J.V.O.).

REFERENCES
[l] M. D. ALA-, B. R. COOKEand A. A. TOPRAC, An investigationof welded open web expanded
beams. Welding Research Supplement, A. W.S. Jl., pp. 775-785 (February 1957).
[2] J. E. GIBBONand W. M. JENKINS,An investigation of the stresses and detiections in castehated beams.
The Structural i%gher 3!5,467-497 (1957).
[3] J. Koroaowsrrt, Stresses and dellections in castellated beams. The Structural Engineer 42.19-24 (1964).
(41 N. J. GARDNW, An investigation into the ddlection behaviour of castellated beams. Trans. EZC-66-Br.
& Sfr. 15 (September 1966).
[S] P. HALLEUX,Etude experimentale et technique du comportement elastique des poutms metailiques a
time evid6e. Revue Francaise de Mechanique, M&moues du Groupement pour 1’Avancement des
Method& d’Analyse des Contraintes, No. 18/19, pp. 123-140 (1966).
[6] P. G. HOWE, Interaction curves for shear and bending of plastic beams. J. Appl. Mech. 24, 453-456
(1957).
[7] J. VANOosrno~, Plastic analysis of the castellated beam. M.A.Sc. Thesis, University of Waterloo, 1969.
[8] D. C. DRUQCER,The effect of shear on the plastic bending of beams. J. Appl. Mech. 23,~514 (1956).
[9] B. G. NEAL, Effect of shear force on the fully plastic moment of an I-Beam. J. Mech. Engng. Sci. 3,
258-266 (1961).
[lo] K. G. E~CKH~FF,The plastic behaviour of sections having one axis of symmetry. Brit. Weld. Res.
Assn. Report, FEI/37 (1954).
Ill] B. G. Nw, The PkzsticMethud of Structural Analyds. Chapman & Hall (1965).
1121 L. S. BEEDLE,Tests of columns under combined thrust and moment. Proc. S.E.S.A. 8,109 (1950).
1131 B. G. NEAL,The effect of shear and normal forces on the fully plastic moment of a beam of rectangular
cross-section. J. Appl. Mech. 28,269-274 (1961).
(141 B. G. NHAL,Effect of shear and normal forces on the fully plastic moment of an I-beam. J. Mech.
Engng. Sci. 3,279-285 (l%l).
(Received 6 January 1971)